The framing effect is a highly robust phenomenon, wherein logically equivalent options (e.g., 90 % chance of winning vs. 10 % chance of losing) trigger different preferences. The Information Leakage account provides a rational interpretation of this effect by suggesting that choice of frame 'leaks' information to decision-makers, making the frames informationally non-equivalent. For example, decision-makers might interpret a positive frame (e.g., 90 % chance of winning) as an implicit recommendation to take a risk. In a series of six preregistered experiments (total N = 1211), we manipulated the informativeness of frames by 1) reducing the perceived freedom of a speaker to choose a frame (the Choice Limitation manipulation), and 2) varying the communication context between the speaker and the listener from collaborative to competitive (the Interest Alignment manipulation). We expected a diminished framing effect in scenarios where the leaked information conveys no useful or trustworthy cues. While the Choice Limitation manipulation occasionally attenuated the framing effect, particularly in within-subject designs, the Interest Alignment manipulation consistently led to a reduction in the framing effect in both within-subject and between-subject designs. These findings show that individuals can be adaptable and sensitive to the informational value of frames and suggest that competition prompts inferences more readily than a speaker's agency over the choice of frame. The implications of these results for rational accounts of framing effects are discussed.